There seems to be some confusion regarding the use of Solutions 2.x graphical analysis tools.

SNR (signal-to-noise) plots are of pretty limited use in analyzing data. The double-difference residual plots are extremely useful. You want to look for trends as well as the magnitude of the residuals. If a given SV shows residuals outside those of the other SV's, removing it will probably improve the solution. It's fairly rare, but if all SV's look very noisy, forbidding that reference SV may improve things. Most often, raising the mask angle or eliminating the worst SV will be the way to go. As far as magnitude goes, allowing for current ionospheric conditions, maybe as much as 0.2 cycles (38mm) wouldn't be unusual. There are several things that can cause unusual trends; ionospheric scintillation, transient iono disturbances, multipath, and it's not always easy (or possible) to distinguish between them. Ideally, the residuals should closely follow the zero line horizontally without sloping or curved trends, and without too may outliers.
It's also important not to overdo data editing, as eliminating too many SV's or raising the mask angle too high can ruin the geometry resulting in a poorer solution.

Hope this helps,
Brian
Modified By Brian D. Ewing, PLS on 8/30/2001 at 9:42 AM

Thanks,

Yes, it is best to look at the double difference carrier phase residuals to see if any cycle slips have not been repaired, as I had mentioned (but maybe not emphasized) in one of my previous posts. Solutions usually repairs cycle slips automatically, but if not there will be a sudden "offset" in the double difference residual plots. Since a cycle slip involves the loss of an integer number of cycles, this offset will be an integer number. Maybe you could call it a spike depending on its width, but the fact remains that it stands out as an outlier from the rest of the

The SNR graph is a first indication of possible problems in the data. When the SNR is lowest, cycle slips are more likely to occur.

In a previous post someone suggested that the mask angle be raised when there are vector failures. This can help sometimes if the problem vector(s) involve satellites that are near the horizon only, and if there are still enough satellites to maintain a quality vector solution. As always, a minimally constrained and constrained network adjustments with independent redundant vectors are good tools to detect biased vectors. This was covered in one of my earlier posts.


Richard

Appreciate the help especially in defining the magnitude we are looking at in terms of normal vs possible/probable error.

I had already come to the conclusion the S/N plot would typically be useful in errors of a large nature ---unknown/unexpected obstructions and the like.


Jimbo